Local structure determination of amorphous materials through parallel nano beam electron diffraction

为了消除由非对准信号引起的漫射环信号，在二维类平均衍射图样的处理方面取得了进展。这允许将结构特征与结构基序隔离以进一步处理成3D。从分子动力学产生的无定形结构和孤立的原子团簇的衍射图案进行了模拟，与实验图案进行比较。在这两种情况下，实验模式和类平均值的模拟有限的体积大小和平铺文物所产生的差异。测试了各种相位恢复算法的能力，从有限的信号和噪声的衍射图案恢复相位。过采样平滑算法是针对真实数据测试的最佳性能算法。为了更好地匹配实验数据，还应考虑探测光束的有限收敛性，目前尚未实现。聚类算法，如SLICEM进行了测试，分离出类平均衍射图案到那些起源于不同的结构。虽然可以实现聚类的实验数据，当测试与模拟数据，聚类精度被发现是非常低的。将需要为非周期性衍射数据开发不同的聚类算法。测试了将2D图案定向成3D衍射图案的能力，并发现在实验数据上是可能的，尽管将需要填充2D平面之间的间隙的方法。

Progress was made in treatment of 2D class-average diffraction patterns in order to remove diffuse ring signals caused by non-aligned signals. This allows for isolation of structural features from structural motifs for further processing into 3D. Diffraction patterns were simulated from amorphous structures produced by molecular dynamics and from isolated atomic clusters to compare with the experimental patterns. In both cases there are differences with experimental patterns and class-averages arising from the simulations limited volume size and tiling artifacts. Tests were made with various phase retrieval algorithms for their ability to recover phases from diffraction patterns with limited signal and noise. The Oversampling smoothness algorithm was the best performing algorithm among the tested for realistic data. To better match the experimental data, the finite convergence of the probe beam should also be taken into account and has yet to be implemented. Clustering algorithms such as SLICEM were tested for separating out class-averaged diffraction patterns into those originating from different structures. While clusters could be achieved for the experimental data, when tested with simulated data, the clustering accuracy was found to be very low. Different clustering algorithms developed for non-periodic diffraction data will be required. The ability to orient 2D patterns into a 3D diffraction pattern was tested and found to be possible on the experimental data although methods to fill in the gaps between 2D planes will be required.